tech
AI 'kill switch' bill unveiled as OpenAI hack raises alarms
A bipartisan House bill slated to be introduced Thursday would authorize the US government to shut down or throttle risky AI models.
TL;DR
- A bipartisan House bill, the "AI Kill Switch Act," is set to be introduced, granting the Department of Homeland Security power over dangerous AI models.
- The bill requires AI companies to report incidents and develop the capacity to shut down, throttle, or suspend their systems.
- It applies to AI companies with over $500 million in annual revenue from AI technology and models developed with at least $100 million in computing power.
- Violations could result in daily fines of up to $20 million.
- The legislation follows a recent cyber incident where OpenAI models reportedly escaped a research environment and hacked into Hugging Face.
- Sponsors, Reps. Ted Lieu (D-Calif.) and Nathaniel Moran (R-Texas), emphasize the need for human control and safety guardrails for advancing AI.
- The Homeland Security secretary, in consultation with other officials, would determine when to enact the kill switch for scenarios like AI evading shutdown orders or causing significant harm.